Portfolio reporting remains one of the last manual processes in many delivery organisations: assembled manually, often late, and open to challenge.

This session shows how Automation Consultants built a 12-agent AI estate to support PMO work across 11 concurrent client engagements, releasing roughly 25% of delivery capacity, and tripling throughput without adding headcount. Working across Jira, Confluence, Loom, and Tempo, with Rovo at the centre, and Anthropic’s Claude as the reasoning layer, the agents support knowledge capture, risk and issue management, budget tracking, status reporting, lessons learnt, resourcing, project provisioning, and proposal drafting.

You’ll see how the system is governed in practice, from the single orchestrating “super agent” that screens every finding, to the Human-in-the-Loop approval gate that keeps people in control before anything reaches Jira, Confluence, or a client inbox. The session will also cover why agent memory lives in state files rather than conversation history, making the setup easier to retire, rebuild, or move between models without losing operational context.

Key takeaways

• How to structure a 12-agent PMO in practice, including which PMO disciplines are safest to automate first.
• Why a single orchestrating agent makes severity, scope, and human approval easier to govern.
• How durable agent memory helps AI workflows survive model, version, and vendor changes.

This is a sponsored session. Badge scanning is optional. By having your conference badge scanned, you are opting in to be contacted by the sponsor. You will be subject to the sponsor’s communications and privacy policies and must contact them directly if you later wish to change your preferences.